The problem of limited availability of data is often faced in engineering. Many situations occur in which the available data is scarce, fragmentary and of poor quality. In this context, uncertainty quantification is a challenging task. In recent years, there has been a growing interest in Bayesian uncertainy quantification methods. This strategy is well suited for parameter estimation when there is an assumed correct model. In this paper, we study the dynamic response of a corrugated skin, a structure that is suitable for applications in aerospace engineering. We define a corrugation unit and model the geometrical parameters as random, incorporating surface roughness. We then produce synthetic, imprecise data and propose a strategy to infer statistics of the dynamic response at a given frequency level subject to a reduced sample of this imprecise data.